We propose a free-energy-perturbation approach accelerated by machine-learning potentials to efficiently compute transition temperatures and entropies for all rungs of Jacob’s ladder. We apply the approach to the dynamically stabilized phases of SiO2, which are characterized by challengingly small transition entropies. All investigated functionals from rungs 1–4 fail to predict an accurate transition temperature by 25–200%. Only by ascending to the fifth rung, within the random phase approximation, an accurate prediction is possible, giving a relative error of 5%. We provide a clear-cut procedure and relevant data to the community for, e.g., developing and evaluating new functionals. 

The VCoNi alloy is a face-centered cubic medium-entropy alloy with exceptionally high yield strength, serving as a prototypical system for investigating short-range order and phase stability in compositionally complex alloys. However, density functional theory calculations underestimate the stability of the random solid solution by several hundred Kelvin. To resolve this discrepancy, we present accurate Gibbs energy calculations for both the random solid solution and a prototypical L12 ordered phase. Our findings reveal that vibrational and electronic excitations account for nearly half of the entropy difference between the ordered phase and disordered solid solution. These factors reduce the energy difference between the two phases by about one-third and help stabilize the solid solution. Our thermodynamic analysis is validated through direct comparison with experimental thermodynamic data.

Machine-learning interatomic potentials (MLIPs) enable large-scale atomistic simulations at moderate computational cost while retaining ab initio accuracy. In recent years, MLIPs trained on coupled-cluster data─particularly CCSD(T), which includes single, double, and perturbative triple excitations─have emerged as a promising route to achieve chemical accuracy (1 kcal/mol) beyond the limits of density functional theory (DFT) and to incorporate nonempirical van der Waals (vdW) interactions. Most existing approaches are, however, still not straightforwardly applicable for systems with extended covalent networks such as covalent organic frameworks (COFs) due to the limited availability of CCSD(T) under periodic boundary conditions. Here we present a methodology to train MLIPs with CCSD(T) accuracy for systems with extended covalent networks. The approach is based on the Δ-learning method with a dispersion-corrected tight-binding baseline and an MLIP trained on the differences of the target CCSD(T) energies from the baseline. This Δ-learning strategy enables training on compact molecular fragments while preserving transferability toward the periodic systems. Dispersion interactions are accounted for by including vdW-bound multimers in the training set, and the combination with a vdW-aware tight-binding baseline allows the formally local MLIP to attain CCSD(T)-level accuracy even for systems dominated by long-range vdW forces. The resulting potential yields root-mean-square energy errors below 0.4 meV/atom on both training and test sets and reproduces electronic total atomization energies, bond lengths, harmonic vibrational frequencies, and intermolecular interaction energies for benchmark molecular systems. We apply the method to a prototypical quasi-two-dimensional covalent organic framework (COF) composed of carbon and hydrogen. The COF structure, interlayer binding energies, and hydrogen absorption are analyzed at CCSD(T) accuracy. Overall, the developed methodology opens a practical route to large-scale atomistic simulations for systems with extended covalent networks and vdW interactions with chemical accuracy.

Oxygen is a commonly overlooked element influencing the properties of many metal oxides. By combining several analytical in situ techniques and theoretical calculations, we demonstrate that oxygen plays a vital part in the phase formation and crystallization of Al2TiO5-based chemical vapor-deposited coatings. Rutherford backscattering spectrometry (RBS) corroborates a polymorphic transformation during crystallization. Subsequent hard X-ray photoelectron spectroscopy (HAXPES) shows that crystallization occurs through a displacive (diffusionless) mechanism. Coupled with theoretical calculations, the crystallization and co-formation of Al2TiO5, Al6Ti2O13, and Al16Ti5O34 are suggested to be driven by the migration of oxygen ions and their corresponding vacancies.

The surface free energies of seven different facets of tungsten (W) are obtained up to the melting point with full account of all the relevant thermal excitations; in particular, thermal atomic vibrations, electronic excitations, and their mutual coupling. The latter is done using ab initio molecular dynamics simulations coupled with the thermodynamic integration technique. In this way, the calculations contain almost no error but the one related to the used exchange-correlation functional, which makes the results truly first principles. The obtained results are compared with previous quasiharmonic calculations for the surface free energies of W and experimental data. The anharmonic contribution is, as expected, important for open surfaces at high temperatures, which leads to a temperature dependence of the surface energy anisotropy. The calculated Wulff shapes and surface energies are in excellent agreement with experimental data close to the melting point, where the crystalline structure of the surface layers is destroyed by a dramatic mobility of the atoms there.

In this work chemical vapour deposited (CVD) coatings of (Tix,W1-x)Ny from TiCl4, WF6, NH3 and Ar were investigated. This coating material has previously been deposited using other vacuum techniques but no publication has so far demonstrated CVD of (Tix,W1-x)Ny. The studied (Tix,W1-x)Ny coatings had a metallic molar ratio (Ti:W) close to 2:1 and 1:1, and were slightly over-stoichiometric with regard to N. The coatings appeared homogeneous and crystallised in a rock salt structure on an alpha-Al2O3 substrate. The cell parameter varied between 4.16 and 4.23 angstrom as a function of the deposition conditions, ranging from a pure TiNx to a pure WNx coating. The texture in the normal direction was (100) for the TiNx and (Tix,W1-x)Ny coatings and (111) for WNx. Electron backscattered diffraction (EBSD) results showed that a strong correlation to the substrate existed but random inplane orientation was also present. The microstructure showed columnar grains with well defined facets growing. Adding a mixture of TiCl4 and WF6 to produce (Tix,W1-x)Ny did increase the grain size significantly when compared to the case when only one metal precursor was present. The down-stream thickness profile, using only WF6 and NH3, displayed mass transport control behaviour, with the coating thickness converging to zero within the deposition zone. Using only TiCl4 on the other hand showed a uniform deposition profile, the signs of a surface kinetics controlled process.

The zero K formation energies of metal and nitrogen vacancies in several (Ti,Al)N alloys and at the (001) (Ti,Al)N/AlN interface are obtained in ab initio supercell calculations. The dependence of the formation energies of metal vacancies on their local environment and type are analyzed and explained in terms of effective cluster interactions for unrelaxed calculations. The common trend for all investigated types of vacancies is that their formation energy increases with the number of Al nearest neighbors if local lattice relaxations are not allowed. However, local lattice relaxations produce a dramatic effect especially in the case of metal vacancies leading to a complicated nonlinear dependence on the local environment indicating the existence of strong multisite strain-induced interactions.

Simulations of spinodal decomposition in an Fe-36 wt%Cr alloy at 773 K are performed by solving the non -linear Cahn-Hilliard equation, and the results are compared with atom probe tomography measurements. The influence of gradient energy coefficient, atomic mobilities and initial structure on the kinetics of spinodal decomposition is studied. It is shown that a proper initial structure, accounting for the thermal history above the miscibility gap, is crucial and enables predictive simulations of spinodal decomposition in Fe-Cr alloys.

This thesis describes the work based on two different tools in computational materials science: a first-principles approach, namely that of density functional theory, and the CALPHAD approach.

These two methods were used in this work to calculate properties of materials related to hard coatings, in particular coatings produced with chemical vapor deposition for the purpose of wear protection in cutting tools. Several parts of the work is also, in many aspects, of a general character. In a few cases, the material investigations were performed on simpler demonstration systems, with the intention of further application on more involved material systems.

A variety of different methods and specific applications are included in this thesis. The reaction-diffusion in Ni-base superalloys deposited by vapor deposition methods was simulated with a continuum approach with CALPHAD thermodynamic and kinetic data. CALPHAD models were also used to predict the stable phases for TiN deposition on a CoCrFeNi substrate. Surfaces and segregation energies were investigated in a random alloy, pseudobinary (Al,Ti)N system. This system was also the subject of calculations of formation energies of structural vacancies, and the configurational dependence of these properties was investigated.

Further, surface free energies including all relevant thermal excitations were calculated for TiN(001) and several W surfaces in a newly developed methodology including machine-learning interatomic potentials. For W, the temperature dependence of the surface anisotropy was obtained, which was shown to be decreasing with temperature, with a surface free energy approaching experimental values at the melting temperature.

Denna avhandling behandlar två olika typer av modellering: modellering från första principer och modellering baserat på CALPHAD-metoden. Dessa två metoder utgör två helt olika tillvägagångssätt i modelleringen av material. För modellering från första principer används täthetsfunktionalteori (eng. förk. DFT), som behandlar elektrondensiteten och baserat på kvantmekanik förutsäger materials egenskaper. CALPHAD är en metod där varje fas i ett materialsystem beskrivs av en termodynamisk tillståndsfunktion som anpassats utifrån en termodynamisk utvärdering där många typer av experimentella data samlas in (och även data från första princip-beräkningar). Generellt modelleras material på en större skala i CALPHAD än med DFT, och CALPHAD-modellerna kopplas ofta till modeller som kan simulera längre längd- och tidsskalor. Första princip-modellering å andra sidan, ger detaljerad insikt i elektronstruktur och atomernas ordnande och rörelser. Denna modellering baseras idealt inte på någon empirisk data alls, utan helt på fysikens grundläggande lagar.Modelleringen i denna avhandling handlar främst om att förbättra förståelsen för de material som beläggs med kemisk ångdeponering (eng. förk. CVD), och i någon mån även beläggningsprocesserna i sig, även om de är mycket komplicerade. Detta inkluderar både modellering av ytor, men också modellering av bulk och gränsskikt. Ur ett makroskopiskt perspektiv är så kallade ytbeläggninar relaterad med ytegenskaper, men många beläggningar för skärande bearbetning är flera mikrometer tjocka, vilket ur ett atomistiskt perspektiv är detsamma som bulken. Det är endast de yttersta nanometrarna som påverkas nämnvärt av en yta.Därför krävs också modellering på flera olika skalor. Vid första princip-beräkningar kan numera, i normala fall, omkring 500 atomer tas med i beräkningen. Det innebär på sin höjd omkring 2–3 nm tjocklek. För att nå större storleksordningar måste andra tekniker användas med lägre precision och vilka kan vara svåra att konstruera, eller som redan nämnts, så kan till exempel CALPHAD användas.I detta arbete har i stort följande områden undersökts: Simuleringav diffusion i en Ni-baslegering under en beläggningsprocess; vakansbildningsenergier, ytenergier och segregeringsenergier i oordnad (Al,Ti)N samt adsorptions- och gränskiktsberäkningar för samma ämne; temperaturberoende för ytenergier hos TiN(001) och olika W-ytor; termodynamiska beräkningar för TiN-beläggning på ett CoCrFeNi-substrat.

The temperature dependence of the surface free energy of the industrially important TiN(001) system has been investigated by means of an extended two-stage upsampled thermodynamic integration using Langevin dynamics (TU-TILD) methodology, to include the fully anharmonic vibrational contribution, as obtained from ab initio molecular dynamics (AIMD). Inclusion of the fully anharmonic behavior is crucial, since the standard low-temperature quasiharmonic approximation exhibits a severe divergence in the surface free energy due to a high-temperature dynamical instability. The anharmonic vibrations compensate for the quasiharmonic divergence and lead to a modest overall temperature effect on the TiN(001) surface free energy, changing it from around 78 meV angstrom(-2) at 0 K to 73 meV angstrom(-2) at 3000 K. The statistical convergence of the molecular dynamics is facilitated by the use of machine-learning potentials, specifically moment tensor potentials, fitted for TiN(001) at finite temperature. The surface free energy obtained directly from the fitted machine-learning potentials is close to that obtained from the full AIMD simulations.